Identifying important structural characteristics ofarsenic resistance proteins by using designedthree-stranded coiled coilsDebra S. Touw*, Christer E. Nordman*, Jeanne A. Stuckey†‡, and Vincent L. Pecoraro*§¶

*Department of Chemistry, †Life Sciences Institute, ‡Biological Chemistry Department, and §Biophysics Research Division, University of Michigan,Ann Arbor, MI 48109

Edited by William F. DeGrado, University of Pennsylvania School of Medicine, Philadelphia, PA, and approved May 30, 2007 (received for reviewMarch 3, 2007)

Arsenic, a contaminant of water supplies worldwide, is one of themost toxic inorganic ions. Despite arsenic’s health impact, there isrelatively little structural detail known about its interactions withproteins. Bacteria such as Escherichia coli have evolved arsenicresistance using the Ars operon that is regulated by ArsR, arepressor protein that dissociates from DNA when As(III) binds. Thisprotein undergoes a critical conformational change upon bindingAs(III) with three cysteine residues. Unfortunately, structures ofArsR with or without As(III) have not been reported. Alternatively,de novo designed peptides can bind As(III) in an endo configurationwithin a thiolate-rich environment consistent with that proposedfor both ArsR and ArsD. We report the structure of the As(III)complex of Coil Ser L9C to a 1.8-Å resolution, providing x-raycharacterization of As(III) in a Tris thiolate protein environment andallowing a structural basis by which to understand arsenated ArsR.

arsenic-binding proteins � coiled coil peptides � crystallography �heavy metal toxicity � protein design

Arsenic toxicity is a worldwide problem as a natural contam-inant of water supplies. Despite the fact that it is a humantoxin and carcinogen, few structural reports on its interactionwith biological ligands have appeared. Escherichia coli and otherbacteria have evolved a detoxification mechanism that employsthe arsRDABC operon (1). Arsenic removal by these encodedproteins is initiated when As(III) binds to ArsR, resulting in thedissociation of the repressor protein from the promoter DNA.The structure of the ArsA component of the ATP-dependentextrusion pump ArsAB with antimonite bound in close proxim-ity to the nucleotide-binding site was able to provide some insightinto the active transport of As(III) out of the cell (2). Addition-ally, structural characterization of substrate and product com-plexes of the arsenate reductase ArsC, which reduces arsenate(AsO43�) to arsenite (AsO2�), has helped elucidate this step ofthe arsenic detoxification pathway (3, 4). Recent studies of E.coli ArsD indicate that it is a metallochaperone, transportingAs(III) to ArsA for extrusion (5). Extended x-ray absorption finestructure and mutagenesis studies have shown that ArsR coor-dinates As(III) with three cysteine thiolates at a distance of 2.25Å, a coordination mode that ArsD is also proposed to employ (1,5). However, no x-ray or NMR structures of ArsR, the repressorprotein, or ArsD have been reported to date. Furthermore, AsS3structures have not been reported for any biologically relevantsmall-molecule thiolates, such as glutathione, and only a handfulof related AsS3 complexes with aromatic ligands or chelated alkyldithiolate coordination are reported (6).

We set out to model the putative As(III) coordination envi-ronments of ArsR and ArsD in a designed peptide system.Previously, we have shown that the three-stranded coiled coilTRI peptides [Ac-G(LaKbAcLdEeEfKg)4G-NH2] designed withthe heptad repeat strategy, such as TRI L16C (sequences of allpeptides are in Table 1), were able to model the unusual trigonalthiolate Hg(II) coordination geometry seen in the MerR me-

talloregulatory protein (7). We have also examined the bindingof the heavy metals Cd(II), Pb(II), and Bi(III) to the TRIpeptides to better understand their modes of toxicity (8, 9).Based on this success, we examined whether the trigonal thiolatemetal-binding cavities of TRI derivatives would be good modelsof the active site of ArsR. As(III) was determined by extendedx-ray absorption fine structure to bind to the peptides TRI L12Cand TRI L16C with As–S bond lengths of 2.24 and 2.25 Å,respectively, as compared with the 2.25-Å As–S distance re-ported for ArsR (10). Armed with this structural similarity, wefelt that an x-ray structure of an arsenated form of the TRIpeptides could reveal important structural insights for the bind-ing of As(III) to ArsR.

The design of coiled coils and other secondary protein foldshas been an active area of research with many structurallycharacterized examples (11, 12). The substantial research incor-porating metal cofactors into designed peptide frameworks hasbeen described in several recent reviews (13–15). Helical scaf-folds have been the predominant structural motif used for theincorporation of many types of metal-binding sites, includinghemes (16), iron–sulfur clusters (17), dinuclear metal centers(18), copper centers (19, 20), and heavy metal-binding trigonalthiolate sites (7, 21). �-Hairpins and other globular structuresthat bind As(III), Zn(II), and other metal cofactors have alsobeen designed (22–25).

Despite the increasing number of reports on the design ofmetallopeptides, detailed structural characterization of many ofthese systems remains elusive. The class of designed metal-lopeptides that has been structurally characterized most thor-

Author contributions: D.S.T., J.A.S., and V.L.P. designed research; D.S.T. performed re-search; D.S.T., C.E.N., J.A.S., and V.L.P. analyzed data; and D.S.T. and V.L.P. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org (PDB ID code 2JGO).

¶To whom correspondence should be addressed. E-mail: vlpec@umich.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0701979104/DC1.

© 2007 by The National Academy of Sciences of the USA

Table 1. Peptide sequences

Peptide gabcdefg abcdefg abcdefg abcdefg

CSL9C Ac-E WEALEKK CAALESK LQALEKK LEALEHG-NH2Coil Ser Ac-E WEALEKK LAALESK LQALEKK LEALEHG-NH2Coil VaLd Ac-G VEALEKK VAALESK VQALESK VEALEHG-NH2TRI Ac-G LKALEEK LKALEEK LKALEEK LKALEEK G-NH2TRI L16C Ac-G LKALEEK LKALEEK CKALEEK LKALEEK G-NH2TRI L12C Ac-G LKALEEK LKACEEK LKALEEK LKALEEK G-NH2

C and N termini are capped by Ac and NH2 groups, respectively.

www.pnas.org�cgi�doi�10.1073�pnas.0701979104 PNAS � July 17, 2007 � vol. 104 � no. 29 � 11969–11974

BIO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

June

1, 2

021

http://www.pnas.org/cgi/content/full/0701979104/DC1http://www.pnas.org/cgi/content/full/0701979104/DC1

oughly is four-helix bundles comprised of dimers of helix–turn–helix peptides. A 2.8-Å resolution structure of a heme-bindingfour-helix bundle maquette has been reported, but by far themost detailed structural information has been obtained for thedue-ferri (DF) family of peptides designed by DeGrado et al.,which contain carboxylate-bridged dinuclear metal centers in thecenter of a four-helix bundle (26, 27). NMR structures of boththe apo and the di-Zn(II) forms of DF1 reveal slight changes inthe protein fold upon metal binding (28). Additionally, a 1.9-Åstructure of a di-Mn(II)-DF2 derivative contains four crystallo-graphically independent bundles that unveil subtle differences inligand coordination to the dinuclear centers (29). The analysis ofthese structures and those of other due-ferri-family derivativesprovides insight into how changes in the hydrophobic core affectligand coordination to these dinuclear sites and has allowed forthe design of catalytic functionality into this type of simpledesigned framework (27, 31).

One of the first designed peptides to be structurally charac-terized was Coil Ser, which was crystallized as an antiparallelthree-stranded coiled coil (32). It was suggested that the anti-parallel orientation was obtained because of the steric bulk of theN-terminal tryptophans. We have been able to show that thereplacement of the interior residue Leu 9 for a cysteine in CoilSer, denoted CSL9C, creates a trigonal thiolate metal-bindingsite within a parallel coiled coil similar to that of TRI L9C (33).CSL9C was shown spectroscopically to have similar behavior toTRI L9C with regard to affinity, pH-dependence, and geometryof metal binding. These results corroborated previous solutionstudies of Coil Ser derivatives suggesting that parallel aggregateswere formed at higher pH values (34). In this study, we are usingCSL9C for structural studies of the arsenic-bound peptidebecause it has proven more facile to form diffraction-qualitycrystals in comparison to the TRI peptides.

The structure presented here of As(CSL9C)3 provides anexamination for As(III) in a homoleptic tristhiolate environmentin a protein. It is precisely this structure that has been proposedfor As(III) binding to ArsR and ArsD. As shown below, there aresubtle, but potentially significant, conformational restrictionsplaced on the As(III) thiolate environment. An analysis of thisstructure not only enhances our understanding of As detoxifi-cation by microbes but, in addition, provides surprising insightsfor the next generation of designed metallopeptides.

ResultsOverall Structure. The crystal structure of the arsenated three-stranded coiled coil Coil Ser L9C As(CSL9C)3 has been deter-mined to a 1.8-Å resolution. Although the sequence of CSL9Cdiffers by only one amino acid from the parent peptide, Coil Ser,the prototypical de novo designed homomeric antiparallel coiledcoil, the structure of As(CSL9C)3 is a well folded, parallelthree-stranded coiled coil similar to the related peptide, CoilVaLd, which has alternating interior hydrophobic layers of valineand leucine residues (35). The final model of As(CSL9C)3contains 762 protein atoms, 50 solvent molecules, 1 As(III) ion,and 4 Zn(II) ions. The rmsd between the search model, which iscomprised of the first 26 amino acids of Coil VaLd, and the finalrefined solution was 1.2 Å. All nonglycine residues fall in themost-favored �-helical region of a Ramachandran plot. TheN-terminal capping acetyl groups have their carbonyl oxygenshydrogen-bonded to alanine 4 of their respective chains. Sidechain residues, with the exception of Glu 24 A, Glu 3 B, and Glu20 B, coordinated to Zn(II) and Glu 6 B, which are involved inhydrogen bonding to a solvent molecule, are in preferredconformations as analyzed by the rotamer check utility in Coot.

The Arsenic-Binding Site. The center of the coiled coil structurecontains an As(III) ion coordinated to the three cysteine resi-dues in a trigonal pyramidal geometry, as seen in Fig. 1.

Refinement of structure with the As(III)–S distance to 2.25 Årestrained to 2.25 � 0.05 Å resulted in a mean As–S bonddistance of 2.28 Å and S–As–S angles of 91°, 92°, and 88°, whichare similar to small molecule As–S3 complexes in the literature(6, 36, 37). The orientation of the leucine residues in the layerabove (Leu 5) and below (Leu 12) the metal-binding site differ,as seen in Fig. 2, with those in the Leu 5 layer, directed moretoward the center of the coiled coil than those in the Leu 12 layer.

The four Zn(II) ions per asymmetric unit present in thestructure lie at crystal-packing interfaces linking the trimerstogether, as seen in Fig. 3. One dinuclear site that is coordinatedby side chains from three separate coiled coils contains a�-(�1,�2)-bridging carboxylate from Glu 24 C that links twoZn(II) ions separated by 4.97 Å. Glu 27 C, His 28 C, Glu 3 C,Glu 24 A, His 28 A, and a water molecule complete the coordi-nation sphere, which is shown in detail in supporting information(SI) Fig. 6A. A second set of Zn(II) ions 4.73 Å apart connectstwo trimers and is coordinated by Glu 1 B, Glu 24 B, Glu 27 B,His 28 B, Glu 6 A, and two water molecules, as can be seen inSI Fig. 6B.

Side Chain Electrostatic Interactions. There are interhelical elec-trostatic interactions between e and g glutamate and lysine sidechains of the heptad repeat on all three interhelical interfaces,which can be observed in detail in SI Fig. 7. There are additionalelectrostatic interactions not typical for a three-stranded coiled

Fig. 1. A top-down view from the N terminus of the coiled coil and a sideview of the trigonal thiolate As(III)-binding site illustrate its pyramidal coor-dination in the interior of the three-stranded coiled coil. (A) Top-down viewof the As(III) ion coordinated to the three cysteine residues in position 9 of theheptad repeat with 2Fo � Fc electron density contoured at 1.5 � overlaid. (B)Side view of the metal-binding site with an Fo� Fc omit map contoured at 5 �demonstrating the endo coordination of the As(III) ion 1.3 Å below the planeof the Cys S�-atoms and on an equal level with the �-methylene protons.

Fig. 2. The packing of the leucine layers above and below the As(III) ion,shown as a purple sphere, is illustrated. (A) Top-down view of the leucine 5layer (blue). (B) Leu 12 layer (orange) looking up the helical axis from the Cterminus; this shows the orientation of the d residues in this layer toward thehelical interface. This illustrates that if a metal that prefers tetrahedral coor-dination such as Cd(II) or Zn(II) were coordinated to the cysteines in the samemanner as As(III), the fourth exogenous ligand (e.g., water) would coordinateon the C-terminal side of the metal ion.

11970 � www.pnas.org�cgi�doi�10.1073�pnas.0701979104 Touw et al.

Dow

nloa

ded

by g

uest

on

June

1, 2

021

http://www.pnas.org/cgi/content/full/0701979104/DC1http://www.pnas.org/cgi/content/full/0701979104/DC1http://www.pnas.org/cgi/content/full/0701979104/DC1http://www.pnas.org/cgi/content/full/0701979104/DC1

coil between the b and g residues Glu 24 A and Lys 22 B andbetween the indole nitrogens of Trp 2 C and Glu 1 A and Trp2 B and Glu 1 C. Several interhelical interactions are mediatedby water molecules, for example, between Glu 1 C and Glu 6 B.In addition, there are several interactions between the backboneand side chains, such as between the hydroxyl group of Ser 14 Cand the carboxyl of the main chain of the same helix at Ala 10.

DiscussionImplications for Arsenic Toxicity. Although arsenic is highly toxicand a carcinogen, its interactions with biomolecules have notbeen thoroughly investigated. Whereas As(V) in the form ofarsenate is thought to bind to phosphate-binding sites (38),As(III) is known to have high affinity for thiolate-rich sites (39).An association constant of 1 � 107 was reported for formationof the trigonal thiolate AsL3 complex with glutathione as mea-sured by calorimetric and spectroscopic methods (39). We expectthat the affinity of As(III) for CSL9C would be greater than orequal to that of glutathione based on our previous studies withthe TRI peptides that demonstrate the effect of peptide self-association on the affinity of Cd(II) and Hg(II) (40). Becauseslow kinetics (3-h incubation at 37°C for glutathione at milli-molar concentration) and low extinction coefficient hinder directdetermination of the affinity of CSL9C for As(III), spectroscopicmeasurement of the displacement of As(III) from As(CSL9C)3by equimolar Cd(II) at pH 8.0 was attempted. This experimentrevealed that As(III) was not replaced by Cd(II) after 5 h,indicating that the Ka of As(III) binding to CSL9C is eithergreater than the value of 2.7 � 107 that was measured for Cd(II)or that the kinetics of As(III) release are too slow to observedisplacement on this time scale (33).

The structure of As(CSL9C)3 reveals a structural example ofAs(III) coordination in a designed peptide environment. Thebest crystallographic model for the As(III) site yields an averageAs–S bond distance of 2.28 Å, which is within the error of thecrystallographic coordinates of our previously determined ex-tended x-ray absorption fine structure values, and an averageS–As–S angle of 90° that are in good agreement with smallmolecule As(III) complexes with trigonal pyramidal Tris-arylthiolate coordination, in which As(III)–S bond distancesranging from 2.23 to 2.29 Å and S–As–S angles with a mean valueof 98° were reported (6, 36, 37). The compression of these angles

in our structure compared with the small molecule may be aresult of peptide aggregation and side chain rotamer effects. Itis very likely that there are similar angles when As(III) binds tothe ArsR and ArsD proteins, with a schematic model for ArsRbeing proposed with As(III)–S distances of 2.25 Å and meanS–As–S 92° angles (1).

This analysis does not address an important conformationalaspect of As(III) binding to a planar three-sulfur environment.The As(III) ion is predicted to have a stereochemically activelone pair. Examination of small-molecule trigonal thiolateAs(III) structures in the literature shows that, in two cases, theAs(III) has a pyramidal geometry with the arsenic located above(exo conformer) the plane of the sulfur atoms of the ligands(based on the As-S-C� atoms of the complexes). In one casewhere the As(III) is oriented below the ring (endo conformer);however, it was thought that this geometry was due to long-rangeoxygen interactions that stabilized the structure (37). Fig. 4shows the endo and exo conformations as they would exist insidea three-stranded coiled coil. Surprisingly, the endo conformationof the As(III) is observed in CSL9C despite the fact that thereare no long-range interactions stabilizing this orientation. Al-though several factors, such as packing of the adjacent hydro-phobes, may be responsible for the observed endo orientation,we believe that it results because the cysteine rotamers areoptimized in this configuration to accommodate the relativelyshort As–S bonds. Previous models for As(III) binding to ArsRhave proposed the arsenic binding to Cys 32, Cys 34, and Cys 37in an �-helical region of the protein. The simultaneous interac-tion of arsenic with Cys 32 and Cys 34 is particularly interestingbecause this is a Cys-X-Cys motif not normally associated withmetal binding and likely requiring significant perturbation of thestructure. Models assessing this perturbation have uniformlyused an As(III)-bound exo to the three cysteines; however, thisstudy demonstrates that future models should consider thepossibility that As(III) adopts an endo conformation. Thus, itmay be useful to reconsider predictions for As(III) coordinationto ArsR and possibly ArsD detoxification proteins.

Influence on Protein Design Strategy. Design of coiled coils hasbeen an active area of research since the report of a paralleltwo-stranded coiled coil by Hodges et al. (41). Since then, designof two-, three-, and four-stranded aggregates has been accom-plished (11). Through the design of coiled coils, it has beendiscovered that the energetics of multiple conformations can bequite similar; this is underscored by the reports of crystal

Fig. 3. The zinc-mediated packing of the coiled coils with Zn(II) and As(III)ions is shown in pink and cyan, respectively. (A) Bottom-up view of a centralcoiled coil (green) with the six Zn(II) ions that are coordinated by side chainsat the C-terminal end. (B) Side view of the trimeric structure of As(CSL9C)3 witha different perspective of the Zn(II) coordination to the exterior residues of thecoiled coils. In both panels, only two of eight symmetry related Zn(II) ions areincluded for clarity.

Fig. 4. Scheme showing the two possible orientations of As(III) within athree-stranded coiled coil. (A) In the exo configuration, the As(III) ion wouldbe located above both S� atoms and �-methylene protons of the cysteine sidechains, and the lone pair would point toward the N terminus. (B) In the endoconfiguration, which is observed in the structure of As(CSL9C)3, the As(III) ionwould be located below the S� atoms and approximately level with the�-methylene protons of the cysteine side chains. The lone pair would then bedirected toward the C terminus.

Touw et al. PNAS � July 17, 2007 � vol. 104 � no. 29 � 11971

BIO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

June

1, 2

021

structures of homomeric assemblies with different aggregationstates (42) and helix orientations (43).

The design of coiled coils that associate specifically intoparallel or antiparallel aggregates has been made possible bymanipulating electrostatic interactions between e and g positionsof the heptad repeat (44, 45). The parallel heteromeric two-stranded coiled coil Velcro peptide and the three-strandedcoiled coil ABC peptide were designed by incorporating repul-sive electrostatic interactions for homomeric species and favor-able electrostatic interactions for heteromeric aggregates (46,47). Additionally, a polar amino acid was incorporated into ahydrophobic core position of the heptad repeat to discourage anantiparallel orientation that would require an unfavorable in-teraction between hydrophobic and polar side chains, whichwould be alleviated by formation of a parallel aggregate. Het-eromeric antiparallel three-stranded coiled coils were designedby using the steric matching of hydrophobic residues to optimizepacking in addition to the incorporation of favorable electro-static interactions (48). In light of this knowledge of the require-ments of electrostatic interactions and hydrophobic packing forthe formation of stable aggregates, it is somewhat surprising thatCoil Ser was found to adopt an antiparallel orientation, in whichtwo of the e–g interhelical interfaces in the trimeric structurecontain pairs of lysine-lysine and glutamate-glutamate residues.Coil Ser has few interhelical interactions in comparison toAs(CSL9C)3. In Coil Ser, there are only four interhelical elec-trostatic interactions between lysine and glutamate residues,between the two helices that are parallel to one another, whereasAs(CSL9C)3 has 10, at all three interfaces, which are illustratedin SI Fig. 7. The pH conditions under which crystals of Coil Serand As(CSL9C)3 were obtained may have an influence on theorientation. Coil Ser was previously crystallized at pH 5.0, wherethe hydrogen bonding capabilities of glutamate residues arepartially minimized by protonation, whereas As(CSL9C)3 wascrystallized at pH 8, where glutamates would be fully deproto-nated. There are also fewer water molecules involved in inter-actions with the peptide Coil Ser, which, in the structurepresented here, are instrumental in interhelical interactions. Theorientation of As(CSL9C)3 as a parallel three-stranded coiledcoil is not surprising, and 2D NMR studies at pH 7.7 indicate aparallel orientation is obtained in solution (O. Iranzo and V.L.P.,unpublished data), even without metal ion coordination tem-plating the orientation (33).

In addition, the cysteine residues, which are in an a position,have their side chains pointed toward the center of the helix. Thechi 1 dihedral angles (N-C�-C-�-S�) for the three cysteineresidues are �60° � 1°, very close to the value of �65° for themost common cysteine rotamer (49). The average crystallo-graphic B-factor for the S� atoms of the cysteine residues is 16,slightly lower than the average B-factor for the entire structure,which is 18. There is no evidence of alternate rotamers for thecysteine side chains. It is possibile that the presence of cysteineresidues at high pH in the center of the coiled coil acts to directthe coiled coil to have a parallel orientation in solution, as wasobserved with polar glutamine and arginine residues in thedesign of parallel heteromeric GCN4 derivatives (50).

In several parallel coiled coil structures of GCN4-derivativepeptides and in Coil VaLd, differences in hydrophobic packing ina and d hydrophobic layers were observed (35, 51, 52). In thesestructures and the one presented here, the hydrophobic layersare composed of either all a or all d residues, and the a leucineresidues tend to orient more toward the center of the coiled coilthan the d residues (35). In the antiparallel Coil Ser structure, theleucine layers are composed of two a and one d� residues or onea and two d� residues, which influences the packing (32).Because hydrophobic residues in a and d positions are known toorient differently with respect to the coiled coil axis, it is likelythat, for a corresponding peptide with cysteines in a d position,

the orientation of the cysteine side chains would be differentfrom those in this case.

One of the original hypotheses as to why Coil Ser adopted anantiparallel orientation was that it did so to avoid steric clashesbetween the three bulky tryptophan residues at position 2 of thesequence (32). Although a reasonable possibility, this structuredemonstrates that there is sufficient room for packing thetryptophan residues at the N terminus, suggesting that this is notthe only influence on the orientation. In addition, in the B andC chains, the tryptophan residues are involved in interhelicalelectrostatic interactions with glutamate side chains. Althoughthe coordination of As(III) to the cysteine residues likely stabi-lizes a parallel trimeric structure, other factors such as electro-static and hydrophobic packing interactions certainly play a rolein determining parallel orientation in solution.

We have successfully designed an As(III) coordination sitewith three cysteines in an endo trigonal pyramidal geometry.Although the observed As–S distances and S–As–S angles arewithin the expected range, the endo configuration was quitesurprising. Our original design incorporated an exo As(III) witha coordination geometry similar to Pb(II) in the protein �-aminolevulinic acid dehydratase (53). This observation raises theinteresting question of whether there are generally structuraldifferences between Pb(II) and As(III) coordination to three-stranded coiled coil peptides and to native proteins. Althoughboth metal ions possess stereochemically active lone pairs andprefer trigonal pyramidal soft ligand-binding sites, As(III) ismuch smaller than Pb(II) (with a typical Pb(II)–S distance of2.64 Å) (9). This difference in size and subsequent S–Pb–S anglesmay not allow Pb(II) to accept coordination as an endo con-former. In addition, the larger size of Pb(II) may perturb thehelix, causing it to pucker out to be accommodated. Another keydistinction between As(III) and Pb(II), other than size, is thedifference in charge between the two ions. It is possible thatnature uses these factors in discriminating between divalent andtrivalent heavy metals in the metalloregulatory systems of theArsR/SmtB family of transcriptional regulator proteins, in whichArsR is used ArsR for detection of As(III) and Sb(III), andCadC is used for detection of Cd(II) and Pb(II) (54).

Analysis of the packing of the Leu residues in the layers aboveand below the As(III)-binding site shows that there is room fora water molecule to bind between the Cys layer at position 9 andLeu 12, as shown in Fig. 2. Using Cd(II), our first designs withthese peptides yielded equilibrium mixtures of three- and four-coordinate Cd(II) (55). By changing the leucine layer above themetal to alanine to allow better water access above the metal-binding site, we designed a peptide (TRI L12AL16C) that bindsCd(II) solely in a four-coordinate geometry, with three cysteinylthiolates and one exogenous water. This observation wouldsuggest that Cd(II) binds to these peptides as the exo conformer,consistent with the longer Cd(II)–S distances (2.54 Å), which aremore reminiscent of Pb(II) than As(III) (8). These observationsmake one consider whether metalloregulatory proteins may beable to differentiate different toxic metals in part by theirpreference for endo [As(III)] versus exo [Pb(II)/Cd(II)] binding.It is known that Sb(III) stimulates the ArsR regulator; thus, itwill be extremely interesting to see whether this ion, which islarger than As(III), will bind in an endo or exo configuration.There is one known structure of ArsA, the membrane translo-cating ATPase, with Sb(III) bound, but it is difficult to interpretits geometrical preferences because three Sb(III) ions are boundin close proximity with bridging cysteine thiolates and chlorideions (2). One Sb(III) ion in this structure has a trigonal cysteinecoordination with two of the cysteines in an endo conformation,an indication that Sb(III) has the capability of binding in thisconformation as well.

Crystallization of this peptide was not successful in theabsence of Zn(II), and the resulting structure shows why it is

11972 � www.pnas.org�cgi�doi�10.1073�pnas.0701979104 Touw et al.

Dow

nloa

ded

by g

uest

on

June

1, 2

021

http://www.pnas.org/cgi/content/full/0701979104/DC1

essential. Zn(II) links the coiled coils together at crystal-packing interfaces on the exterior of the peptide by coordi-nation to glutamate and histidine side chains. The binding ofZn(II) by the side chains of the C-terminal residues results inthe C terminus of CSL9C being better folded in this region incomparison to Coil VaLd. A least-squares overlay of the C�from residues 3 through 18 has an average rmsd of 0.2 Å fromthat of VaLd; however, if the entire helix is included, the rmsdis 1.7 Å, demonstrating the greater differences at the N and Ctermini, as can be seen in Fig. 5. It is likely that Zn(II) may

facilitate the crystallization of other Coil Ser derivative pep-tides (e.g., CSL19C) that have been shown to have differentmetal-binding properties than CSL9C (33).

ConclusionsWe present here a unique structure of a designed three-strandedcoiled coil peptide with As(III) bound to the hydrophobic interiorin a trigonal thiolate coordination geometry. It is bound slightlybelow the plane of the cysteine residues in an endo trigonalpyramidal geometry. This structure provides a good model of theactive site of the ArsR repressor protein when it is bound to As(III),which has not yet been structurally characterized. In addition, thediscovery of Zn(II) ions linking the coiled coils together in the solidstate is an interesting and potentially useful observation for thecrystallization of other designed coiled coil peptides.

Experimental ProceduresPeptide Synthesis and Purification. CSL9C was synthesized on anApplied Biosystems (Foster City, CA) 433A peptide synthesizer bystandard protocols and purified and characterized as previouslyreported (33). As(CSL9C)3 was prepared by incubating 10-foldexcess NaAsO2 with peptide in 50 mM potassium phosphate buffer(pH 7.0). The solution was allowed to react at room temperature for36–48 h and then purified by RP-HPLC on a linear water toacetonitrile gradient. The lyophilized off-white powder was char-acterized by MALDI-MS in a sinapinic acid matrix to have theexpected mass of 10,076 Da for the As(III)–peptide complex.

Crystallization. As(CSL9C)3 was dissolved in distilled Milli-Qwater to a concentration of 15 mg/ml. Crystals were grown byvapor diffusion at 20°C in a sitting drop with equal volumes ofpeptide and precipitant containing 100 mM imidazole buffer(pH 8.0), 200 mM Zn(OAc)2, and 40% PEG 400. Crystals werefrozen in their mother liquor for data collection.

Data Collection and Refinement. Data were measured at theAdvanced Photon Source (APS) (COM-CAT Beamline 32-IDand GMCA-CAT Beamline 23-ID) at the Argonne NationalLaboratory and were collected on a MarCCD (Mar USA,Evanston, IL) at a wavelength of 1.00 Å at �180°C. Threehundred sixty frames of data were collected with a 1° rotationand 1-s exposure. Data were processed and scaled with theprogram HKL-2000 (56). The crystals were of the space groupC2 with unit cell parameters a � 77.28, b � 44.20, c � 29.41, � �� � 90°, and � � 119.5°.

The structure was solved by molecular replacement, using as amodel the trimeric coiled coil Coil VaLd (Protein Data Bank entry1COI) with the chain truncated to 26 residues. The side chains ofthe model were included. The Patterson-based program used wasan updated version of GENPAT (35, 57). This model with sidechains truncated to alanine was refined by rigid body refinementand restrained refinement in Refmac 5 in the CCP4 suite ofprograms (58, 59). 2Fo � Fc and Fo � Fc electron density mapsgenerated with the CCP4 map utilities were used for building in sidechains, metal ions, and C-terminal residues in Coot. Coot was usedfor finding water molecules (60). The refinement to 1.8 Å resultedin Rworking � 19.8% and Rfree � 25.4%. Data collection andrefinement statistics are given in Table 2. The validity of the modelwas verified by a composite omit map generated in the Crystallog-raphy and NMR System (30). Figures were generated in Pymol andMIFit. The coordinates and structure factors have been depositedin the Protein Data Bank with the ID code 2JGO.

We thank Joseph Brunzelle for data collection. This research wassupported by National Institutes of Health Grant ES012236 (to V.L.P.),the Chemical Biology Interface Training Grant (to D.S.T.), and fundsfrom the Michigan Economic Development Corporation Life SciencesCorridor (to J.A.S.).

Table 2. Data collection and refinement statistics

Data collection statistics

Data set NativeSpace group C2Unit cell a � 77.28; b � 29.41;

c � 44.19; � � 90; � � 119.5; � � 90Wavelength, Å 1.00Resolution, Å 1.8 (1.8–1.86)Rsym, % 6.5 (20.3)�I/�I� 10 (3)Completeness, % 98.4 (90.6)Redundancy 7 (5)

Refinement statistics

Resolution, Å 1.8 (13.7–1.8)R-factor, % 19.8Rfree, % 25.4Protein atoms 762Water molecules 50Unique reflections 8,137rmsd

Bonds 0.018Angles 1.448

Parenthesized numbers are the values for the highest resolution shell.

Fig. 5. The overlay of Coil VaLd, shown in green (PDB entry 1COI), a relatedparallel three-stranded coiled coil with As(CSL9C)3 (red) demonstratestheir structural similarity and highlights their divergence at the C and Ntermini where the Zn(II) ions hold CSL9C in a more helical conformationthan VaLd.

Touw et al. PNAS � July 17, 2007 � vol. 104 � no. 29 � 11973

BIO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

June

1, 2

021

1. Shi W, Dong J, Scott RA, Ksenzenko MY, Rosen BP (1996) J Biol Chem271:9291–9297.

2. Zhou T, Radaev S, Rosen BP, Gatti DL (2000) EMBO J 19:4838–4845.3. Martin P, DeMel S, Shi J, Gladysheva T, Gatti DL, Rosen BP, Edwards BFP

(2001) Structure (London) 9:1071–1081.4. Messens J, Martins JC, Van Belle K, Brosens E, Desmyter A, De Gieter M,

Wieruszeski J-M, Willem R, Wyns L, Zegers I (2002) Proc Natl Acad Sci USA99:8506–8511.

5. Lin Y-F, Walmsley AR, Rosen BP (2006) Proc Natl Acad Sci USA 103:15617–15622.

6. Shaikh TA, Bakus RC, II, Parkin S, Atwood DA (2006) J Organomet Chem691:1825–1833.

7. Dieckmann GR, McRorie DK, Tierney DL, Utschig LM, Singer CP,O’Halloran TV, Penner-Hahn JE, DeGrado WF, Pecoraro VL (1997) J AmChem Soc 119:6195–6196.

8. Matzapetakis M, Farrer BT, Weng T-C, Hemmingsen L, Penner-Hahn JE,Pecoraro VL (2002) J Am Chem Soc 124:8042–8054.

9. Matzapetakis M, Ghosh D, Weng T-C, Penner-Hahn JE, Pecoraro VL (2006)J Biol Inorg Chem 11:876–890.

10. Farrer BT, McClure CP, Penner-Hahn JE, Pecoraro VL (2000) Inorg Chem39:5422–5423.

11. Woolfson DN (2005) in Fibrous Proteins: Coiled-Coils, Collagen and Elastomers,ed Squire JM (Academic, San Diego), pp 79–112.

12. Pantoja-Uceda D, Santiveri CM, Jiménez MA (2006) in Protein Design—Methods and Applications, Methods in Molecular Biology, eds Guerois R, dela Paz ML (Humana, Totowa, NJ), Vol 340, pp 27–52.

13. Doerr AJ, McLendon GL (2004) Inorg Chem 43:7916–7925.14. Ghosh D, Pecoraro VL (2005) Curr Opin Chem Biol 9:97–103.15. Barker PD (2003) Curr Opin Struct Biol 13:490–499.16. Reedy CJ, Kennedy ML, Gibney BR (2003) Chem Commun, 570–571.17. Gibney BR, Mulholland SE, Rabanal F, Dutton PL (1996) Proc Natl Acad Sci

USA 93:15041–15046.18. Marsh ENG, DeGrado WF (2002) Proc Natl Acad Sci USA 99:5150–5154.19. Schnepf R, Haehnel W, Wieghardt K, Hildebrandt P (2004) J Am Chem Soc

126:14389–14399.20. Kharenko OA, Kennedy DC, Demeler B, Maroney MJ, Ogawa MY (2005)

J Am Chem Soc 127:7678–7679.21. Li X, Suzuki K, Kanaori K, Tajima K, Kashiwada A, Hiroaki H, Kohda D,

Tanaka T (2000) Protein Sci 9:1327–1333.22. Petros AK, Reddi AR, Kennedy ML, Hyslop AG, Gibney BR (2006) Inorg

Chem 45:9941–9958.23. Ramadan D, Cline DJ, Bai S, Thorpe C, Schneider JP (2007) J Am Chem Soc

129:2981–2988.24. Daugherty RG, Wasowicz T, Gibney BR, DeRose VJ (2002) Inorg Chem

41:2623–2632.25. Mulholland SE, Gibney BR, Rabanal F, Dutton PL (1998) J Am Chem Soc

120:10296–10302.26. Huang SS, Gibney BR, Stayrook SE, Leslie Dutton P, Lewis M (2003) J Mol

Biol 326:1219–1225.27. Lombardi A, Summa CM, Geremia S, Randaccio L, Pavone V, DeGrado WF

(2000) Proc Natl Acad Sci USA 97:6298–6305.

28. Maglio O, Nastri F, Pavone V, Lombardi A, DeGrado WF (2003) Proc NatlAcad Sci USA 100:3772–3777.

29. DeGrado WF, Costanzo LD, Geremia S, Lombardi A, Pavone V, RandaccioL (2003) Angew Chem Int Ed 42:417–420.

30. Brunger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-KunstleveRW, Jiang J-S, Kuszewski J, Nilges M, Pannu NS, et al. (1998) Acta CrystallogrD 54:905–921.

31. Kaplan J, DeGrado WF (2004) Proc Natl Acad Sci USA 101:11566–11570.32. Lovejoy B, Choe S, Cascio D, McRorie D, DeGrado W, Eisenberg D (1993)

Science 259:1288–1293.33. Iranzo O, Ghosh D, Pecoraro VL (2006) Inorg Chem 45:9959–9973.34. Wendt H, Berger C, Baici A, Thomas RM, Bosshard HR (1995) Biochemistry

34:4097–4107.35. Ogihara NL, Weiss MS, DeGrado WF, Eisenberg D (1997) Protein Sci 6:80–88.36. Shaikh TA, Parkin S, Atwood DA (2006) J Organomet Chem 691:4167–4171.37. Carter TG, Healey ER, Pitt MA, Johnson DW (2005) Inorg Chem 44:9634–

9636.38. Willsky GR, Malamy MH (1980) J Bacteriol 144:356–365.39. Spuches AM, Kruszyna HG, Rich AM, Wilcox DE (2005) Inorg Chem

44:2964–2972.40. Ghosh D, Lee K-H, Demeler B, Pecoraro VL (2005) Biochemistry 44:10732–

10740.41. Hodges R, Saund A, Chong P, St.-Pierre S, Reid R (1981) J Biol Chem

256:1214–1224.42. Gonzalez L, Brown RA, Richardson D, Alber T (1996) Nat Struct Biol

3:1002–1010.43. Yadav MK, Leman LJ, Price DJ, Brooks CL, Stout CD, Ghadiri MR (2006)

Biochemistry 45:4463–4473.44. Monera OD, Kay CM, Hodges RS (1994) Biochemistry 33:3862–3871.45. Schnarr NA, Kennan AJ (2005) Org Lett 7:395–398.46. O’Shea EK, Lumb KJ, Kim PS (1993) Curr Biol 3:658–667.47. Nautiyal S, Woolfson DN, King DS, Alber T (1995) Biochemistry 34:11645–

11651.48. Schnarr NA, Kennan AJ (2004) J Am Chem Soc 126:14447–14451.49. Ponder JW, Richards FM (1987) J Mol Biol 193:775–791.50. Oakley MG, Hollenbeck JJ (2001) Curr Opin Struct Biol 11:450–457.51. Harbury PB, Zhang T, Kim PS, Alber T (1993) Science 262:1401–1407.52. Harbury PB, Kim PS, Alber T (1994) Nature 371:80–83.53. Erskine PT, Duke EMH, Tickle IJ, Senior NM, Warren MJ, Cooper JB (2000)

Acta Crystallogr D 56:421–430.54. Pennella MA, Giedroc DP (2005) BioMetals 18:413–428.55. Lee K-H, Cabello C, Hemmingsen L, Marsh ENG, Pecoraro VL (2006) Angew

Chem Int Ed 45:2864–2868.56. Otwinowski Z, Minor W (1997) Methods Enzymol 276:307–326.57. Nordman C (1994) Acta Crystallogr A 50:68–72.58. Storoni LC, McCoy AJ, Read RJ (2004) Acta Crystallogr D 60:432–438.59. Collaborative Computational Project Number 4 (1994) Acta Crystallogr D

50:760–763.60. Emsley P, Cowtan K (2004) Acta Crystallogr D 60:2126–2132.

11974 � www.pnas.org�cgi�doi�10.1073�pnas.0701979104 Touw et al.

Dow

nloa

ded

by g

uest

on

June

1, 2

021